15-441 Computer Networking
Lecture 7 - Ethernet
Peter Steenkiste
Departments of Computer Science and
Electrical and Computer Engineering
15-441 Networking, Spring 2008
http://www.cs.cmu.edu/~dga/15-441/S08
1
Problem: Sharing a Wire
Learned how to connect hosts

… But what if we want more hosts?

Wires for everybody!
Switches
Expensive! How can we share a wire?
2
Listen and Talk
yak yak…

Natural scheme – listen before you talk…
»Works well in practice
3
3
Listen and Talk
yada yada…

Natural scheme – listen before you talk…
»Works well in practice
4
4
Listen and Talk
yada
yak yak…
yada…

Natural scheme – listen before you talk…
»Works well in practice

But sometimes this breaks down
»Why? How do we fix/prevent this?
5
5
Problem: Who is this packet for?

Need to put an address on the packet

What should it look like?

How do you determine your own address?

How do you know what address you want to send it
to?
6
6
Outline

Aloha

Ethernet MAC

Collisions

Ethernet Frames
7
7
Random Access Protocols

When node has packet to send
» Transmit at full channel data rate R
» No a priori coordination among nodes
Two or more transmitting nodes  “collision”
 Random access MAC protocol specifies:

» How to detect collisions
» How to recover from collisions (e.g., via delayed
retransmissions)

Examples of random access MAC protocols:
» Slotted ALOHA and ALOHA
» CSMA and CSMA/CD
8
8
Aloha – Basic Technique

First random MAC developed
» For radio-based communication in Hawaii (1970)

Basic idea:
» When you are ready, transmit
» Receivers send ACK for data
» Detect collisions by timing out for ACK
» Recover from collision by trying after random
delay
– Too short  large number of collisions
– Too long  underutilization
9
9
Slotted Aloha

Time is divided into equal size slots
» Equal to packet transmission time

Node (w/ packet) transmits at beginning of next slot

If collision: retransmit pkt in future slots with
probability p, until successful
Success (S), Collision (C), Empty (E) slots
10
10
Pure (Unslotted) ALOHA

Unslotted Aloha: simpler, no synchronization

Pkt needs transmission:
» Send without awaiting for beginning of slot

Collision probability increases:
» Pkt sent at t0 collide with other pkts sent in [t0-1, t0+1]
11
11
Outline

Aloha

Ethernet MAC

Collisions

Ethernet Frames
12
12
Ethernet
First practical local area network, built at
Xerox PARC in 70’s
 “Dominant” LAN technology:

» Cheap
» Kept up with speed race: 10, 100, 1000 Mbps
Metcalfe’s Ethernet
sketch
13
13
Ethernet MAC – Carrier Sense

Basic idea:
» Listen to wire before
transmission
» Avoid collision with
active transmission

Why didn’t ALOHA
have this?
» In wireless, relevant
contention at the
receiver, not sender
Hidden
Exposed
St.Louis
NY
Chicago
CMU
CMU
Chicago
NY
– Hidden terminal
– Exposed terminal
14
14
Ethernet MAC – Collision
Detection

But: ALOHA has collision detection also?
» That was very slow and inefficient

Basic idea:
» Listen while transmitting
» If you notice interference  assume collision

Why didn’t ALOHA have this?
» Very difficult for radios to listen and transmit
» Signal strength is reduced by distance for radio
– Much easier to hear “local, powerful” radio station
than one in NY
– You may not notice any “interference”
15
15
Ethernet MAC (CSMA/CD)

Carrier Sense Multiple Access/Collision
Detection
Packet?
No
Sense
Carrier
Send
Detect
Collision
Yes
Discard
Packet
attempts < 16
Jam channel
b=CalcBackoff();
wait(b);
attempts++;
attempts == 16
16
16
Ethernet CSMA/CD:
Making it word
Jam Signal: make sure all other transmitters
are aware of collision; 48 bits;
Exponential Backoff:
 If deterministic delay after collision,
collision will occur again in lockstep
 Why not random delay with fixed mean?
» Few senders  needless waiting
» Too many senders  too many collisions

Goal: adapt retransmission attempts to
estimated current load
» heavy load: random wait will be longer
17
17
Ethernet Backoff Calculation

Exponentially increasing random delay
»Infer senders from # of collisions
»More senders  increase wait time



First collision: choose K from {0,1}; delay is K
x 512 bit transmission times
After second collision: choose K from
{0,1,2,3}…
After ten or more collisions, choose K from
{0,1,2,3,4,…,1023}
18
18
Outline

Aloha

Ethernet MAC

Collisions

Ethernet Frames
19
19
Collisions
B
C
Time
A
20
20
Minimum Packet Size

What if two people
sent really small
packets
» How do you find
collision?
21
21
Ethernet Collision Detect

Min packet length > 2x max prop delay
»If A, B are at opposite sides of link,
and B starts one link prop delay after
A

Jam network for 32-48 bits after collision,
then stop sending
»Ensures that everyone notices
collision
22
22
End to End Delay

c in cable = 60% * c in vacuum = 1.8 x 10^8
m/s

Modern 10Mb Ethernet {
» 2.5km, 10Mbps
» ~= 12.5us delay
» +Introduced repeaters (max 5 segments)
» Worst case – 51.2us round trip time!

Slot time = 51.2us = 512bits in flight
» After this amount, sender is guaranteed sole
access to link
» 51.2us = slot time for backoff
23
23
Packet Size

What about scaling? 3Mbit, 100Mbit,
1Gbit...
» Original 3Mbit Ethernet did not have minimum
packet size  bonus question!
– Max length = 1Km and No repeaters
» For higher speeds must make network smaller,
minimum packet size larger or both

What about a maximum packet size?
» Needed to prevent node from hogging the
network
» 1500 bytes in Ethernet
24
24
10BaseT and 100BaseT

10/100 Mbps rate; latter called “fast ethernet”

T stands for Twisted Pair (wiring)

Minimum packet size requirement
» Make network smaller  solution for 100BaseT
25
25
Gbit Ethernet

Minimum packet size requirement
» Make network smaller?
– 512bits @ 1Gbps = 512ns
– 512ns * 1.8 * 10^8 = 92meters = too small !!
» Make min pkt size larger!
– Gigabit Ethernet uses collision extension for small pkts
and backward compatibility

Maximum packet size requirement
» 1500 bytes is not really “hogging” the network
» Defines “jumbo frames” (9000 bytes) for higher
efficiency
26
26
Outline

Aloha

Ethernet MAC

Collisions

Ethernet Frames
27
27
Ethernet Frame Structure

Sending adapter encapsulates IP datagram (or other
network layer protocol packet) in Ethernet frame
28
28
Ethernet Frame Structure (cont.)

Preamble: 8 bytes
»101010…1011
»Used to synchronize receiver,
sender clock rates

CRC: 4 bytes
»Checked at receiver, if error is
detected, the frame is simply
dropped
29
29
Ethernet Frame Structure (cont.)

Each protocol layer needs to provide
some hooks to upper layer protocols
» Demultiplexing: identify which upper layer
protocol packet belongs to
» E.g., port numbers allow TCP/UDP to identify
target application
» Ethernet uses Type field

Type: 2 bytes
» Indicates the higher layer protocol, mostly IP
but others may be supported such as Novell
IPX and AppleTalk)
30
30
Addressing Alternatives

Broadcast  all nodes receive all packets
» Addressing determines which packets are kept and
which are packets are thrown away
» Packets can be sent to:
– Unicast – one destination
– Multicast – group of nodes (e.g. “everyone playing Quake”)
– Broadcast – everybody on wire

Dynamic addresses (e.g. Appletalk)
» Pick an address at random
» Broadcast “is anyone using address XX?”
» If yes, repeat

Static address (e.g. Ethernet)
31
31
Ethernet Frame Structure (cont.)

Addresses: 6 bytes
» Each adapter is given a globally unique
address at manufacturing time
– Address space is allocated to manufacturers

24 bits identify manufacturer

E.g., 0:0:15:*  3com adapter
– Frame is received by all adapters on a LAN and
dropped if address does not match
» Special addresses
– Broadcast – FF:FF:FF:FF:FF:FF is “everybody”
– Range of addresses allocated to multicast

Adapter maintains list of multicast groups node is
interested in
32
32
Why Did Ethernet Win?

Failure modes
» Token rings – network unusable
» Ethernet – node detached

Good performance in common case
» Deals well with bursty traffic
» Usually used at low load

Volume  lower cost  higher volume ….

Adaptable
» To higher bandwidths (vs. FDDI)
» To switching (vs. ATM)

Easy incremental deployment

Cheap cabling, etc
33
33
And .. It is Easy to Manage

You plug in the host and it basically works
» No configuration at the datalink layer
» Today: may need to deal with security
Protocol is fully distributed
 Broadcast-based.

» In part explains the easy management
» Some of the LAN protocols (e.g. ARP) rely on
broadcast
– Networking would be harder without ARP
» Not having natural broadcast capabilities adds
complexity to a LAN
– Example: ATM
34
34
Summary

CSMA/CD  carrier sense multiple access
with collision detection
» Why do we need exponential backoff?
» Why does collision happen?
» Why do we need a minimum packet size?
– How does this scale with speed?

Ethernet
» What is the purpose of different header fields?
» What do Ethernet addresses look like?

What are some alternatives to Ethernet
design?
37
37
Lecture 6
Datalink – Framing, Switching
Peter Steenkiste
Departments of Computer Science and
Electrical and Computer Engineering
Carnegie Mellon University
15-441 Networking, Spring 2008
http://www.cs.cmu.edu/~dga/15-441/S08
45
From Signals to Packets
Analog Signal
“Digital” Signal
Bit Stream
Packets
0 0 1 0 1 1 1 0 0 0 1
0100010101011100101010101011101110000001111010101110101010101101011010111001
Header/Body
Packet
Transmission
Sender
Header/Body
Header/Body
Receiver
46
Datalink Functions

Framing: encapsulating a network layer
datagram into a bit stream.
» Add header, mark and detect frame boundaries, …

Media access: controlling which frame should
be sent over the link next.
» Easy for point-to-point links; half versus full duplex
» Harder for multi-access links: who gets to send?

Error control: error detection and correction
to deal with bit errors.
» May also include other reliability support, e.g.
retransmission

Flow control: avoid that the sender outruns
the receiver.
47
Datalink Lectures



Framing and error coding.
Datalink architectures.
Switch-based networks.
» Packet forwarding
» Flow and error control

Taking turn protocols.


Contention-based networks: basic Ethernet.
Ethernet bridging and switching.
Connectivity to the home.

Circuit-based communication

48
Framing



A link layer function, defining which bits have
which function.
Minimal functionality: mark the beginning and
end of packets (or frames).
Some techniques:
»
»
»
»
out of band delimiters (e.g. FDDI 4B/5B control symbols)
frame delimiter characters with character stuffing
frame delimiter codes with bit stuffing
synchronous transmission (e.g. SONET)
49
Character and Bit Stuffing

Mark frames with special character.
» What happens when the user sends this character?
» Use escape character when controls appear in data:
*abc*def -> *abc\*def
» Very common on serial lines, in editors, etc.

Mark frames with special bit sequence
» must ensure data containing this sequence can be
transmitted
» example: suppose 11111111 is a special sequence.
» transmitter inserts a 0 when this appears in the data:
» 11111111 -> 111111101
» must stuff a zero any time seven 1s appear:
» 11111110 -> 111111100
» receiver unstuffs.
50
Example: Ethernet Framing
preamble


datagram
length
more stuff
Preamble is 7 bytes of 10101010 (5 MHz
square wave) followed by one byte of
10101011
Allows receivers to recognize start of
transmission after idle channel
51
SONET



SONET is the Synchronous Optical Network
standard for data transport over optical fiber.
One of the design goals was to be backwards
compatible with many older telco standards.
Beside minimal framing functionality, it
provides many other functions:
» operation, administration and maintenance (OAM)
communications
» synchronization
» multiplexing of lower rate signals
» multiplexing for higher rates
52
Standardization History

Process was started by divestiture in 1984.
» Multiple telephone companies building their own
infrastructure




SONET concepts originally developed by
Bellcore.
First standardized by ANSI T1X1 group for the
US.
Later picked up by CCITT and developed its
own version.
SONET/SDH standards approved in 1988.
53
A Word about Data Rates


Bandwidth of telephone channel is under
4KHz, so when digitizing:
8000 samples/sec * 8 bits = 64Kbits/second
Common data rates supported by telcos in
North America:
» Modem: rate improved over the years
» T1/DS1: 24 voice channels plus 1 bit per sample
(24 * 8 + 1) * 8000 = 1.544 Mbits/second
» T3/DS3: 28 T1 channels:
7 * 4 * 1.544 = 44.736 Mbits/second
54
Synchronous Data Transfer

Sender and receiver are always synchronized.
» Frame boundaries are recognized based on the clock
» No need to continuously look for special bit sequences

SONET frames contain room for control and data.
» Data frame multiplexes bytes from many users
» Control provides information on data, management, …
3 cols
transport
overhead
87 cols payload capacity
9 rows
55
SONET Framing

Base channel is STS-1 (Synchronous Transport System).
» Takes 125 sec and corresponds to 51.84 Mbps
» 1 byte/frame corresponds to a 64 Kbs channel (voice)
» Transmitted on an OC-1 optical carrier (fiber link)

Standard ways of supporting slower and faster channels.
» Support both old standards and future (higher) data rates

Actual payload frame “floats” in the synchronous frame.
» Clocks on individual links do not have to be synchronized
3 cols
transport
overhead
87 cols payload capacity,
including 1 col path overhead
9 rows
56
How Do We Support
Lower Rates?

125 sec
1 Byte in every consecutive
frame corresponds to a 64
Kbit/second channel.
» 1 voice call.

125 sec
Higher bandwidth channels
hold more bytes per frame.
» Multiples of 64 Kbit/second

Channels have a “telecom”
flavor.
125 sec
» Fixed bandwidth
» Just data – no headers
» SONET multiplexers remember
how bytes on one link should be
mapped to bytes on the next link
– Byte 33 on incoming link 1
is byte 97 on outgoing link 7
57
How Do We Support
Higher Rates?

125 sec

Send multiple frames in a 125
sec time slot.
The properties of a channel
using a single byte/ST-1
frame are maintained!


125 sec
» Constant 64 Kbit/second rate
» Nice spacing of the byte samples
125 sec
Rates typically go up by a
factor of 4.
Two ways of doing
interleaving.
» Frame interleaving
» Column interleaving
– concatenated version, i.e.
OC-3c
58
The SONET Signal Hierarchy
STS-1 carries
one DS-3 plus
overhead
Signal Type
line rate
# of DS0
DS0 (POTS)
DS1
DS3
64 Kbs
1.544 Mbs
44.736 Mbs
1
24
672
OC-1
OC-3
OC-12
STS-48
STS-192
51.84 Mbs
155 Mbs
622 Mbs
2.49 Gbs
9.95 Gbs
672
2,016
8,064
32,256
129,024
STS-768
39.8 Gbs
516,096
59
Using SONET in Networks
Add-drop capability allows soft configuration of networks,
usually managed manually.
OC-48
mux
DS1
OC-3c
mux
OC-12c
mux
60
Self-Healing SONET Rings
OC-48
mux
mux
mux
mux
DS1
OC-3c
OC-12c
61
SONET as Physical Layer
OC3/12
Access
OC3/12
Access CO
OC12/48
Metro
CO
POP
OC3/12
Access
CO
CO
OC3/12
Access
OC12/48
Metro
CO
WDM Backbone
OC48/192
POP
OC3/12
Access
POP
OC3/12
Access
CO
OC12/48
Metro
CO
OC3/12
Access
62
Error Coding

Transmission process may introduce errors
into a message.
» Single bit errors versus burst errors

Detection:
» Requires a convention that some messages are invalid
» Hence requires extra bits
» An (n,k) code has codewords of n bits with k data bits
and r = (n-k) redundant check bits

Correction
» Forward error correction: many related code words map
to the same data word
» Detect errors and retry transmission
63
Basic Concept:
Hamming Distance



Hamming distance of two
bit strings = number of bit
positions in which they
differ.
If the valid words of a code
have minimum Hamming
distance D, then D-1 bit
errors can be detected.
If the valid words of a code
have minimum Hamming
distance D, then [(D-1)/2] bit
errors can be corrected.
1 0 1 1 0
1 1 0 1 0
HD=2
HD=3
64
Examples

A (4,3) parity code has D=2:
0001 0010 0100 0111 1000 1011 1101 1110
(last bit is binary sum of previous 3, inverted - “odd parity”)

A (7,4) code with D=3 (2ED, 1EC):
0000000
0100011
1000110
1100101



0001101
0101110
1001011
1101000
0010111
0110100
1010001
1110010
0011010
0111001
1011100
1111111
1001111 corrects to 1001011
Note the inherent risk in correction; consider
a 2-bit error resulting in 1001011 -> 1111011.
There are formulas to calculate the number of
extra bits that are needed for a certain D.
65
Cyclic Redundancy Codes
(CRC)

Commonly used codes that have good error
detection properties.
» Can catch many error combinations with a small number or
redundant bits

Based on division of polynomials.
» Errors can be viewed as adding terms to the polynomial
» Should be unlikely that the division will still work


Can be implemented very efficiently in
hardware.
Examples:
» CRC-32: Ethernet
» CRC-8, CRC-10, CRC-32: ATM
66
Datalink Architectures


Packet forwarding.
Error and flow control.


Media access
control.
Scalability.
67
Media Access Control


How do we transfer packets between two hosts
connected to the same network?
Switches connected by point-to-point links -store-and-forward.
» Used in WAN, LAN, and for home connections
» Conceptually similar to “routing”
– But at the datalink layer instead of the network layer
» Today

Multiple access networks -- contention based.
»
»
»
»
Multiple hosts are sharing the same transmission medium
Used in LANs and wireless
Need to control access to the medium
Mostly Thursday lecture
68
A Switch-based Network


Switches are connected by point-point links.
Packets are forwarded hop-by-hop by the switches
towards the destination.
» Forwarding is based on the address



How does a switch work?
How do nodes exchange packets over a link?
How is the destination addressed?
Switch
PCs at
Work
Point-Point
link
PC at
Home
69
Switching Introduction


Idea: forward units of data based on address in
header.
Many data-link technologies use switching.
» Virtual circuits: Frame Relay, ATM, X.25, ..
» Packets: Ethernet, MPLS, …

“Switching” also happens at the network layer.
»
»
»
»

Layer 3: Internet protocol
In this case, address is an IP address
IP over SONET, IP over ATM, ..
Otherwise, operation is very similar
Switching is different from SONET mux/demux.
» SONET channels statically configured - no addresses
70
An Inter-network
Host
Host
Host
Host
Ethernet
Host
Host
Wireless
Ethernet
Framerelay
Host
Host
Host
Host
Ethernet
ATM
Host
IP/SONET
Host
802.X
Host
Host
Host
71
Internetworking Options
7
6
5
4
3
2
1
physical
1
7
6
5
4
3
2
1
repeater
7
6
5
4
3
2
1
network
3
2
2
1
1
router
7
6
5
4
3
2
1
data link
2
1
1
7
6
5
4
3
2
1
Switching/bridging
(e.g. 802 MAC)
7
6
5
4
3
2
1
7
6
5
4
3
2
1
...
3
2
1
3
2
1
7
6
5
4
3
2
1
gateway
72
Switch Architecture

Takes in packets in one
interface and has to forward
them to an output interface
based on the address.
» A big intersection
» Same idea for bridges, switches,
routers: address look up differs

Control processor manages
the switch and executes
higher level protocols.
» E.g. routing, management, ..


The switch fabric directs the
traffic to the right output port.
The input and output ports
deal with transmission and
reception of packets.
Control
Processor
Input
Port
Output
Port
Output
Port
Input
Port
Switch
Fabric
Output
Port
Input
Port
Output
Port
Input
Port
73
Packet Forwarding:
Address Lookup
Switch
Address Next Hop
B31123812508
38913C3C2137
A21023C90590
3
3
0
Info
13
-

» Absolute address (e.g. Ethernet)
» (IP address for routers)
» (VC identifier, e.g. ATM))


128.2.15.3
1
(2,34)
Address from header.

Next hop: output port for packet.
Info: priority, VC id, ..
Table is filled in by routing protocol.
74
Link Flow Control and
Error Control




Naïve protocol.
Dealing with receiver overflow: flow control.
Dealing with packet loss and corruption: error control.
Meta-comment: these issues are relevant at many
layers.
» Link layer: sender and receiver attached to the same “wire”
» End-to-end: transmission control protocol (TCP) - sender and
receiver are the end points of a connection

How can we implement flow control?
» “You may send” (windows, stop-and-wait, etc.)
» “Please shut up” (source quench, 802.3x pause frames, etc.)
» Where are each of these appropriate?
75
A Naïve Protocol


Sender simply sends to the receiver whenever it has
packets.
Potential problem: sender can outrun the receiver.
» Receiver too slow, buffer overflow, ..

Not always a problem: receiver might be fast enough.
Sender
Receiver
76
Adding Flow Control

Stop and wait flow control: sender waits to send the
next packet until the previous packet has been
acknowledged by the receiver.
» Receiver can pace the receiver

Drawbacks: adds overheads, slowdown for long links.
Sender
Receiver
77
Window Flow Control


Stop and wait flow control results in poor throughput
for long-delay paths: packet size/ roundtrip-time.
Solution: receiver provides sender with a window that
it can fill with packets.
» The window is backed up by buffer space on receiver
» Receiver acknowledges the a packet every time a packet is
consumed and a buffer is freed
Sender
Receiver
78
Bandwidth-Delay Product
RTT
Sender
Receiver
Time
Max Throughput =
Window Size
Roundtrip Time
79
Dealing with Errors
Stop and Wait Case

Packets can get lost, corrupted, or duplicated.
» Error detection or correction turns corrupted packet in lost or
correct packet


Duplicate packet: use sequence numbers.
Lost packet: time outs and acknowledgements.
» Positive versus negative acknowledgements
» Sender side versus receiver side timeouts

Window based flow control: more aggressive use of
sequence numbers (see transport lectures).
Sender
Receiver
80
What is Used in Practice?

No flow or error control.
» E.g. regular Ethernet, just uses CRC for error detection

Flow control only.
» E.g. Gigabit Ethernet

Flow and error control.
» E.g. X.25 (older connection-based service at 64 Kbs that
guarantees reliable in order delivery of data)
81
Datalink Layer Architectures


Packet forwarding.
Error and flow control.


Media access
control.
Scalability.
82
Datalink Classification
Datalink
Switch-based
Virtual
Circuits
Packet
Switching
ATM,
framerelay
Bridged
LANs
Multiple Access
Scheduled
Access
Random
Access
Token ring,
Ethernet,
FDDI, 802.11 802.11, Aloha
83
83
Multiple Access Protocols

Prevent two or more nodes from transmitting
at the same time over a broadcast channel.
» If they do, we have a collision, and receivers will not be
able to interpret the signal

Several classes of multiple access protocols.
» Partitioning the channel, e.g. frequency-division or time
division multiplexing
– With fixed partitioning of bandwidth –
– Not flexible; inefficient for bursty traffic
» Taking turns, e.g. token-based, reservation-based
protocols, polling based
» Contention based protocols, e.g. Aloha, Ethernet
– Next lecture
84
Fiber Distributed Data Interface
(FDDI)

One token holder may send,
with a time limit
» Provides known upper bound on
delay.




Optical version of 802.5 token
ring, but multiple packets may
travel in train: token released
at end of frame
100 Mbps, 100km
Optional dual ring for fault
tolerance
Concerns:
» Token overhead
» Latency
» Single point of failure
85
Other “Taking Turn”
Protocols

Central entity polls stations, inviting them to
transmit
» Simple design – no conflicts
» Not very efficient – overhead of polling operation
» Example: the “Point Control Function” mode for 802.11

Stations reserve a slot for transmission.
» For example, break up the transmission time in
contention-based and reservation based slots
– Contention based slots can be used for short
messages or to reserve time slots
– Communication in reservation based slots only
allowed after a reservation is made
» Issues: fairness, efficiency
86
MAC Protocols - Discussion



Channel partitioning MAC protocols:
» Share channel efficiently at high load
» Inefficient at low load: delay in channel
access, 1/N bandwidth allocated even if
only 1 active node!
“Taking turns” protocols
» More flexible bandwidth allocation, but
» Protocol can introduce unnecessary
overhead and access delay at low load
Random access MAC protocols (next lecture)
» Efficient at low load: single node can fully
utilize channel
» High load: collision overhead
Lecture 7: 9-19-06
87
87